Platforms

The infrastructure you'll actually live in.

Model & code hubs

• Hugging Face: huggingface.co. The GitHub of ML: 1M+ models, datasets, Spaces (live demos), and the transformers/datasets libraries. A center of gravity for open-source AI. 🔄
• GitHub Trending: github.com/trending. Filter by day or week to catch breakout AI repos before the newsletters do. 🔄
• Ollama: ollama.com. Run open models locally with one command; an easy on-ramp to local inference.
• LM Studio: lmstudio.ai. A desktop app for downloading and running local models with a GUI; the point-and-click peer to Ollama.
• Replicate: replicate.com. Run and fine-tune open models via API without managing GPUs. Useful for prototyping.
• Together / Fireworks / Groq: together.ai · fireworks.ai · groq.com. Fast, cheap hosted inference for open models (Groq for very low latency). 🔄
• OpenRouter: openrouter.ai/models. A single API and marketplace across 400+ hosted models, with live per-token pricing and usage rankings. 🔄
• llms.txt: llmstxt.org. Emerging convention, a Markdown file that points models at your site's key docs/content to make them AI-readable. Worth adopting for docs and products. 🔄

Benchmarks & leaderboards (Papers-with-Code successors)

• LMArena (formerly Chatbot Arena; the older lmarena.ai now redirects here): arena.ai. Human-preference Elo rankings; the closest thing to a "vibes benchmark" that people trust. 🔄
• Artificial Analysis: artificialanalysis.ai. Independent index comparing models on intelligence, speed, and price across providers; widely cited for head-to-head numbers. 🔄
• Hugging Face Open LLM ecosystem & leaderboards: huggingface.co/spaces. After the original Open LLM Leaderboard was retired, evaluation moved to task-specific Spaces and Arena-style boards. Search Spaces for the current one for your task. 🔄
• Papers With Code: paperswithcode.com. Largely archived and frozen now. Still a useful historical map of benchmark SOTA. Its living successors are alphaXiv and HF Papers (below).
• alphaXiv: alphaxiv.org. Community discussion layer over arXiv; a practical Papers-with-Code successor for tracking active work. 🔄
• Hugging Face Daily Papers: huggingface.co/papers. Curated, upvoted daily paper feed. A fast triage tool. 🔄

Dataset sources

• Hugging Face Datasets: huggingface.co/datasets. The default hub for training and eval data.
• Kaggle: kaggle.com/datasets. Datasets, competitions, and free notebooks/GPUs; a good learning sandbox.
• Common Crawl: commoncrawl.org. The web-scale corpus underlying most pretraining.
• Awesome Public Datasets: github.com/awesomedata/awesome-public-datasets. Curated index across many domains.